Meta AI Open-Sources Rebalancer: A C++ Assignment Solver That Runs About 40 Million Placement Problems a Day
40 million placement problems a day. Meta just open-sourced the C++ solver that keeps its infrastructure running.
Why it matters
Meta has released Rebalancer, a production-grade assignment solver used internally for 9+ years to optimize shard, server, and traffic placement at scale. The open-source release (Apache 2.0, pip-installable) gives enterprises and operators access to the same algorithmic infrastructure Meta uses to solve combinatorial optimization at data-center scale—relevant for teams building or scaling distributed systems and AI compute clusters.
The key facts
10 to know40 million assignment problems solved daily
9+ years of production deployment at Meta
Supports local search and MIP solvers: Gurobi, FICO Xpress, HiGHS
C++ and Python, Apache 2.0 license
Pip-installable for immediate adoption
Rebalancer handles ~40 million assignment problems per day at Meta
9+ years of internal deployment history
Apache 2.0 license, pip-installable
C++ and Python libraries
Solves shard placement, server allocation, and traffic rebalancing
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Meta has open-sourced Rebalancer, the C++ and Python library it has used for over 9 years to place shards, servers and traffic. It handles about 40 million assignment problems a day, using local search or MIP solvers like Gurobi, FICO Xpress and HiGHS. It is pip-installable under Apache 2.0.